Singularity Matrix-S

(5)

In Stock

| >10 in Storage

2799€*

* +19% VAT and Delivery Costs

A specialized compute card designed for AI research and machine learning development, not for gaming. It features high-density tensor cores for accelerating neural network training.

Technical Data

Boost Clock	1800 MHz
Memory Size	32 GB
Memory Type	HBM3e
Memory Bus	5120-bit
Memory Bandwidth	3277 GB/s
Compute Units	8,192
Tdp	400 W (Passive, for Server Airflow)
Power Connectors	1x 16-pin
Outputs	none
Card Length	267 mm

Product Description

The Singularity Matrix-S is a purpose-built accelerator for AI workloads. It eschews traditional graphics outputs and ray-tracing hardware in favor of 8,192 fourth-generation tensor cores optimized for FP16, bfloat16, and INT8 computations. The card includes 32GB of HBM3e memory, providing extreme bandwidth for training large language models and complex datasets. It connects via a PCIe 5.0 x16 interface and supports direct GPU-to-GPU communication over a high-speed bridge. Cooling is handled by a server-grade passive heatsink designed for high-airflow rack environments.

Technical Data

Boost Clock	1800 MHz
Memory Size	32 GB
Memory Type	HBM3e
Memory Bus	5120-bit
Memory Bandwidth	3277 GB/s
Compute Units	8,192
Tdp	400 W (Passive, for Server Airflow)
Power Connectors	1x 16-pin
Outputs	none
Card Length	267 mm

Product Description

The Singularity Matrix-S is a purpose-built accelerator for AI workloads. It eschews traditional graphics outputs and ray-tracing hardware in favor of 8,192 fourth-generation tensor cores optimized for FP16, bfloat16, and INT8 computations. The card includes 32GB of HBM3e memory, providing extreme bandwidth for training large language models and complex datasets. It connects via a PCIe 5.0 x16 interface and supports direct GPU-to-GPU communication over a high-speed bridge. Cooling is handled by a server-grade passive heatsink designed for high-airflow rack environments.